An Efficient Text Analyzer by Prosody Generator Driven Approach for Mandarin TTS

• Main Authors: Cheng-Yu Yeh, 葉政育
• Other Authors: Shaw-Hwa Hwang
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/78289254266232939531

碩士 === 國立臺北科技大學 === 電機工程系碩士班 === 90 === An efficient text analyzer which is driven by prosody generator for Mandarin text-to-speech will be analyzed and designed in this thesis. Three research topics are analyzed in detail and the system implementation is done efficiently according to these analytic results. For the topic #1, the contribution of each linguistic feature on the prosody generator will be analyzed. For the topic #2, the cross-influence of each linguistic feature on the prosody generator will be analyzed. For the last topic, the entropy of linguistic feature will be employed to examine the problem of over-classification on the linguistic feature. Finally, the efficient text analyzer architecture will be designed based on these analytic results. For the performance testing, totally 39,103 Chinese characters can be segmented into word sequences within 78ms under personal computer(P4-1.4G.). The correction rate of word tagging with 97% is achieved. It confirms that the performance of our text analyzer is very good.